Reference: MLFW-2707
Tensorboard API is attached to the tensorflow SDK while in reality it could work for Pytorch as well. We would need to make estimator not required in the api:
https://github.com/aws/sagemaker-python-sdk/blob/0e97997c88101b7f5603f0a8657e6f6a94d2784e/src/sagemaker/tensorflow/estimator.py#L56
and pass the path to the logdir we would write to in the local filesystem.
Hello @tanguycdls,
Thank you for the awesome suggestion. I'll add this into our backlog, which we are always reprioritizing based on customer feedback.
Reference: MLFW-2707
Thanks!
This would be so useful! I would value this feature immensely to improving my workflow and efficiency.
Most helpful comment
Hello @tanguycdls,
Thank you for the awesome suggestion. I'll add this into our backlog, which we are always reprioritizing based on customer feedback.
Reference: MLFW-2707
Thanks!